GSA inks $1 Anthropic AI deal for US government, expands access
The General Services Administration (GSA) has finalized another significant agreement in its push to integrate artificial intelligence across the U.S. federal government, announcing a deal that makes Anthropic’s Claude AI available to agencies for a mere dollar per year. This latest “OneGov” initiative, unveiled on August 12, 2025, extends access to Anthropic’s Enterprise and Government variants of Claude across all three branches of government. The OneGov framework streamlines procurement, allowing agencies to adopt Claude products without the need for individual purchasing agreements with Anthropic.
Josh Gruenbaum, Commissioner of the GSA’s Federal Acquisition Service, hailed the agreement as a “major step in the AI-driven transformation of government,” asserting that OneGov is “revolutionizing how the federal government acquires AI technology.” As part of the arrangement, Anthropic will also provide essential support to government agencies, assisting with onboarding processes. Notably, Claude’s Government offering holds FedRAMP High certification, indicating it meets the most rigorous security standards for handling sensitive, unclassified government workloads.
A key aspect of this deal is its potential reach beyond federal civilian executive agencies. While awaiting official confirmation from the GSA or Anthropic, it appears to mark the first instance in a series of recent AI procurement agreements where discounted AI access is extended to members of Congress and the federal judiciary. Their participation, however, remains contingent on the approval of their respective branches, suggesting a novel expansion of AI availability within government.
Yet, this expansion, particularly to the federal judiciary, raises significant concerns. The legal community has been grappling with an increasing number of instances where AI-generated content includes fabricated legal citations – a phenomenon often termed “hallucinations.” This issue becomes particularly salient given Anthropic’s own past. The company’s lawyers were previously called out in court for using fabricated legal citations in a music copyright case, with the error attributed to Claude’s “hallucinations.” Providing an AI tool with such a history to judges, who rely on factual accuracy for their rulings, prompts serious questions about its suitability and potential risks.
This agreement with Anthropic is the latest in a series of aggressive AI adoption initiatives by the GSA. It follows closely on the heels of another major announcement just a week prior, when the agency added Claude, Gemini, and ChatGPT to its Multiple Award Schedule, further streamlining their purchase by agencies under the OneGov strategy. On the same day, the GSA also announced a nearly identical $1 per year deal with OpenAI, making ChatGPT available to federal agencies, complete with support.
The GSA’s strategy extends beyond AI models to broader technology infrastructure. The agency previously struck deals with Amazon Web Services (AWS), providing federal agencies with $1 billion in AWS credits through the end of 2028, and with Oracle, which offers license-based technology at a 75 percent discount. However, a critical characteristic shared by all these deals is their temporary nature. The AWS credits expire, Oracle’s discounts are not permanent, and the $1 AI deals with OpenAI and Anthropic are set to conclude after a year.
This temporary pricing structure presents a potential challenge for federal agencies. As they integrate and become accustomed to using these premium services at heavily discounted rates, the expiration of these trial periods could lead to significant “vendor lock-in.” Agencies might find themselves compelled to pay substantially higher prices for software that has become integral to their operations. Inquiries to the GSA regarding their plans to mitigate these potential lock-in issues for federal agencies have so far gone unanswered.